Deleterious nnutations in BRCA1 and BRCA2 increase ovarian cancer risk substantially, but such mutations are rare in the population and collectively they account for fewer than 15 percent of cases. GWAS have successfully identified many common susceptibility alleles for multiple disease phenotypes, including the identification of genes involved in pathways not previously implicated in cancer. This has not been possible for ovarian cancer, but the recent/imminent completion of several independent GWAS with the potential for replication in participating OCAC studies provides the opportunity to achieve this and match the success for other cancers, in Project 1, the first aim is tol pool data from four GWAS that used similar genotyping platforms, use several different approaches to analyze the data, and follow-up the most promising associations (SNPs and CNVs) in a large-scale replication in independent data sets. Genotypes are available from the lllumina HumanHap 610 chip on roughly 4400 cases and 4200 controls through UK and US GWAS, on 500 cases and 300 controls using the lllumina 317k chip, and a pooled DNA analysis of 520 cases and 900 controls in Australia which employed the lllumina 1M beadtype chip. In addition, the UK GWAS has genotyped nearly 22,000 promising SNPs on an additional roughly 5,000 cases and 5,000 controls;and the US GWAS will interrogate 16,000 SNPs on an additional 6,000 cases and 6,000 controls in October 2009. The second aim will be to replicate the findings using a custom iSelect lllumina chip of 13,680 beadtypes on a Stage III study population of an additional roughly 3,500 cases and 3,500 controls that have not been included in any of the previous Stage I or Stage II genotyping efforts. The total estimated threestage sample size will be approximately 16,400 cases and 17,100 controls, providing excellent power to detect risk alleles, critical to Projects 2 and 3. The third aim is to examine the association of SNPs and CNVs from all pooled Stage I and Stage II data in relation to ovarian cancer survival. For aim four we will design and run a custom iSelect lllumina chip of 3,072 SNPs on a Stage HI study population of additional cases for whom overall survival data are available. The total estimated three-stage survival sample size will be approximately 13,700 cases. We will fit ordinal genetic models to predict overall survival accounting for known clinical prognostic factors and conduct exploratory analysis of SNP associations within other major histologic subtypes (e.g., mucinous and clear cell).